A Computational Method of Inferring Disease Diffusion Networks and its Empirical Validation in Epidemiologic Surveillance

The aim of this project is to develop and empirically evaluate a computational method that allows for inferring and predicting transmission patterns in outbreaks of emerging infectious diseases (e.g., H5N1 and SARS) as well as in resurgence of endemic diseases (e.g., malaria) in populations. The spread of infectious diseases can be caused and affected by multiple (or even hidden) factors, which makes accurate prediction of when and where outbreaks may emerge difficult. At the moment, temporal-spatial scan statistics-based clustering techniques have been applied to the analysis and characterization of temporal-spatial patterns of infectious diseases. However, the existing clustering-based surveillance software systems do not infer the underlying diffusion patterns (modeled as networks) of infectious diseases, which can be informative and perceptive as they characterize how diseases are diffused or transmitted from one location to another over time. By integrating both temporal-spatial clusters of cases of infection and temporal-spatial diffusion networks of diseases, existing regional or national surveillance systems can further enhance their functional capacities of predicting/analyzing the impact of disease transmissions and their underlying factors, which can benefit healthcare providers, government infectious disease control and prevention organizations (e.g., Hong Kong Department of Health - Centre for Health Protection (CHP)), as well as international agencies (e.g., World Health Organization (WHO)).

Technically, the problem of computationally inferring diffusion networks of infectious diseases is both interesting and challenging because, during the process of disease spread, the interrelationships between reported infection cases are not observed directly; what we have is only raw surveillance data on temporal-spatial distributions of infection cases. Moreover, we need to incorporate different disease propagation models, because different infectious diseases can have different infectious spread mechanisms. In this project, we will develop and demonstrate a computational method to automatically infer the underlying diffusion network of an infectious disease that incorporates (1) surveillance data, i.e., the temporal-spatial distributions of cases of infection, and (2) infection/propagation models of the disease. In addition, we will conduct empirical validations of the proposed method as well as its application to real-world case studies by implementing a prototype and testing it in modeling disease transmission patterns based on available surveillance datasets (e.g., in the practical cases of malaria and respiratory disease surveillance).